Multi-Trip Multi-Compartment Vehicle Routing

Problem with Backhauls

Sukhpal Ramanand
University of Delhi
KAUSHAL KUMAR (  [email protected] )
University of Delhi https://orcid.org/0000-0001-9313-5657

Research Article

Keywords: Heterogeneous Fleet, Multi-Compartment, Multi-Trip, Pickup and Delivery, Vehicle Routing
Problem, Backhauls

Posted Date: April 7th, 2023

DOI: https://doi.org/10.21203/rs.3.rs-2761446/v1

License:   This work is licensed under a Creative Commons Attribution 4.0 International License.
Read Full License
Multi-Trip Multi-Compartment Vehicle Routing
Problem with Backhauls
Sukhpal* and Kaushal Kumar*
Department of Operational Research, Faculty of Mathematical Sciences, University of Delhi, Delhi, India
*Corresponding author. Email: [email protected]; [email protected]

Abstract
The Multi-Trip Multi-Compartment Vehicle Routing Problem with Backhauls (MTMCVRPB) is a
complex optimization problem that involves finding the most efficient routes for a heterogeneous fleet
of multi-compartment vehicles to transport heterogeneous commodities simultaneously. The problem
involves making deliveries and pickups while taking into account capacity constraints on compartments
and other specific requirements such as limitation on route length. The MTMCVRPB aims to minimize
the total cost of the routes, including the cost of vehicles, fuel and other expenses, while maximizing
efficiency of vehicles and fulfilling the demand of customers. In order to solve the MTMCVRPB, this
article recommends a mixed integer linear programming (MILP) formulation. The proposed model takes
into consideration the demands of both linehaul and backhaul customers, ensuring that on each route
linehaul deliveries are made before visiting any backhaul customer. The model was tested on benchmark
instances from literature and randomly generated data sets and numerical experiments were performed to
demonstrate its effectiveness. The proposed model can be beneficial for the industries such as grocery and
food delivery, fuel distribution and garbage collection as these are main applications of this model. We
have also provided a saving based heuristic by making certain modifications to find the quick and efficient
solution to the proposed model.

Keywords: Heterogeneous Fleet, Multi-Compartment, Multi-Trip, Pickup and Delivery, Vehicle Routing Problem,
Backhauls

1. INTRODUCTION
The vehicle routing problem (VRP) is a combinatorial optimization that entails determining the best
route for a fleet of vehicles to traverse in order to serve a number of customers. The researchers G.B.
Dantzig and J.H. Ramser firstly generalized the traveling salesman problem as the truck dispatching
problem in their article (Dantzig and Ramser 1959). They provided the first mathematical program-
ming model and algorithmic approach as well as a description of a practical application involving
the distribution of fuel to service stations. Numerous models and techniques for the exact and ideal
solution of the different VRP variants have been proposed after the publication of this landmark study
and new VRP variants have been researched in order to better align with the real-world applications.

An enhanced version of the method developed by Dantzig and Ramser as an iterative process
to solve the above mentioned truck dispatching problem was presented in (Clarke and Wright 1964)
and popularly known as "Clarke and Wright’s Saving Algorithm". This saving algorithm is used to
solve both the basic and more complex versions of vehicle routing problems.

The VRP is an NP-hard problem. So, there may be a limit to the size of problems that can be solved
optimally using mathematical programming or combinatorial optimization. Because of the size
and frequency of real-world VRPs, commercial solvers frequently adhere to heuristics. For simple
tasks, exact algorithms are employed. But, we know that real-world problems are very large in
scale. Hence, heuristic and metaheuristic algorithms are stronger alternatives to solve larger problems.

In logistics, distribution and transportation, the vehicle routing problem is the most researched
combinatorial optimization problem. There is humongous literature on vehicle routing problems
e.g. (Toth and Vigo 2002), (Toth and Vigo 2014) and (Mor and Speranza 2022) analyzed more than
thousands of the VRPs. In order to analyze the gaps, we examined the three kinds of vehicle routing
problems, MTVRP (Şen and Bülbül 2008), VRPB (Santos et al. 2020) and the third one on which
we mainly focused is MCVRP (Ostermeier et al. 2021). To the best of our knowledge, no work
exists that concurrently combines the characteristics of all these three aspects. These VRPs described
briefly ahead.

MTVRP
The extension of the basic vehicle routing problem that involves numerous trips being made by the
same vehicle in a predetermined time period is known as the multi-trip vehicle routing problem
(MTVRP). Vehicle routing problems with this variation were initially investigated in (Fleischmann
1990). Following that, other researchers made contributions to this category of vehicle routing
problems like (Brandão and Mercer 1998), (Mingozzi, Roberti, and Toth 2013) and (Cattaruzza,
Absi, and Feillet 2016). In the people transportation sector and many other industries, the demand of
a large number of customers is satisfied with vehicles that possess a limited capacity. Vehicles are
employed repeatedly over time to maximise the profitability and to make the best use of the vehicles.
In prior literature on MTVRPs (Şen and Bülbül 2008) a variety of MTVRPs with various properties
are explored, along with numerous formulations and solution techniques.

VRPB
The category known as "vehicle routing problem with backhauls" (VRPB) involves two different set
of customers. First set is known as linehauls, each customer in this set requires a given quantity to be
delivered from depot and second set is called backhauls, each backhaul customer requires a given
quantity to be picked up by vehicles. This is a variant of pickup and delivery vehicle routing problem
(PDVRP) in which all delivery customers (linehauls) should be satisfied prior to picking up from
any pickup point (backhaul) by the vehicle on a route. The significance of the linehaul-backhaul
problem is related to the ongoing effort to reduce distribution costs by utilizing the unused capacity
of an empty vehicle returning to the depot. The initial studies (Golden et al. 1985) and (Goetschalckx
and Jacobs-Blecha 1989) discussed the vehicle routing problem with backhauls.

MCVRP
All delivered goods in several industries are homogeneous in the sense that they can be transported
together. Other industries, on the other hand, have inhomogeneous goods. To reduce transportation
costs, those industries frequently use vehicles with compartments to transport heterogeneous goods
in different storage containers on the same vehicle. In multi-compartment vehicle routing problems,
vehicles with loaded space or containers divided into a number of compartments and are utilized to
collect goods with various characteristics or to distribute several commodities being carried together.
The use of multi-compartment vehicles enable the joint delivery of various types of products from a
depot to the customers on a single vehicle. For instance (Chajakis and Guignard 2003), trucks that
transport liquid fuels have separated tanks that can hold fuels of various varieties, such as gasoline of
different grades and maritime boats that transport petroleum products from refineries to clients have
numerous compartments that can carry more than five distinct goods simultaneously in a single trip.

Problem addressed in this article is known as multi-trip multi-compartment vehicle routing problem
 3

with backhauls (MTMCVRPB) since we merged the three aforementioned categories. In this class
of VRP a multi-compartment vehicle performs numerous trips to fulfill the demands of pickup and
delivery customers. In each trip, vehicle first delivers goods to all linehauls(delivery points), if the
multi-commodity requirements of every linehaul customer on the vehicle’s route have been met, it
then picks up goods from backhauls(pickup locations) before returning to the depot.

The remainder of this work is structured as follows: The relevant literature is covered in section
2. In section 3, we have described the problem and the underlying assumptions. The MILP of the
given problem is provided in section 4. The solution approach based on Clarke and Wright’s saving
algorithm is mentioned in section 5. A validation of the proposed model is provided in section 6.
Section 7 presents conclusion.

2. LITERATURE REVIEW
The associated literature is covered in this section. We have mostly concentrated on research per-
taining to multi-compartment vehicle routing problems. The Multi-Compartment Vehicle Routing
Problem (MCVRP) is a variant of the classical Vehicle Routing Problem (VRP) that considers vehicles
with multiple compartments for carrying different products. The MCVRP adds an extra layer of
complexity to the VRP by requiring the consideration of capacity constraints for each compartment
of the vehicle. Literature on the MCVRP dates back to the 1980s, with numerous studies focusing on
various aspects of the problem, such as solution methods, optimization algorithms and problem varia-
tions. The majority of these studies have attempted to address the additional constraints introduced by
the multiple compartments, such as limited capacity, budget constraint and the need for product com-
patibility. One of the most widely used methods for solving the MCVRP is the use of metaheuristics,
such as genetic algorithms and simulated annealing (see table 1). These algorithms have been shown
to be effective in finding near-optimal solutions for large-scale MCVRP instances in a reasonable
amount of time. Recent research has also focused on developing exact algorithms for the MCVRP,
such as branch-and-cut and branch-and-price methods. These methods have been successful in
solving smaller MCVRP instances to optimality, but their scalability remains a challenge for larger
problems. In addition, researchers have also investigated problem variations, such as the MCVRP with
time windows, multiple depots and heterogeneous vehicles. These variations add further constraints
and complexities to the MCVRP, requiring the development of new solution methods and algorithms.

These multi-compartment vehicles first time addressed in (Chajakis and Guignard 2003) for the
transportation of foods to convenience stores. This study aims to reduce cooling and delivery costs.
The costs associated with maintaining non-ambient temperature commodities at the proper tem-
peratures during deliveries are known as cooling costs. The 0-1 programming and mixed integer
programming mathematical models were developed and was solved using the Lagrangean Relaxations.
On the subject of applications, models and heuristics for vehicle routing with compartments, (Derigs
et al. 2011) gave a good overview and summary. An integer programming model of the MCVRP
was provided in this paper as a model formulation. Many heuristics and metaheuristics are discussed
as well as comparisons of the solutions of benchmark instances. Other literatures on the modeling,
optimizations, heuristics and metaheuristics can also be found in (Coelho and Laporte 2015), (Koch,
Henke, and Wäscher 2016), (Kurnia et al. 2018) and (Reil, Bortfeldt, and Mönch 2018).

In the study (Asawarungsaengkul et al. 2013), the reason for using vehicles with multiple com-
partments is to transport products of varying temperatures or compositions in the same vehicle. They
addressed the concept in two applications that make use of vehicles with multiple compartments
which are, petroleum products are transported, groceries and foods are delivered to convenience
stores. In the case of a fleet of vehicles operating from a single depot, food and groceries are delivered
to a set of customers. Customer orders are divided into three categories: frozen, refrigerated and
dry. The MCVRP can be used in many real-world applications where the goods are different in
characteristics, such as milk (cows and goats) and fluid product (gasoline and diesel), different in
quality, such as expired dates, different in treatment conditions, such as level of storage temperatures
and so on. Vehicles with multiple compartments are used to save transportation costs by transporting
heterogeneous goods in different compartments on the same vehicle.

Several studies in the literature have utilized various real-world cases to address different varia-
tions of the Multi-Compartment Vehicle Routing Problem (MCVRP). The study (Ostermeier and
Hübner 2018), (Hübner and Ostermeier 2019), (Martins et al. 2019) and (Frank et al. 2021) utilized
grocery product-based cases to solve the problem. On the other hand, the studies (Chen, Liu, and
Langevin 2019), (Chen and Shi 2019), (Kim and Park 2020) and (Chen, Pan, and Yi 2022) used
temperature-based multi-compartment vehicles to address the issue. Manufacturing-based examples
are discussed in (Pasha et al. 2020) and (Zou, Pan, and Tasgetiren 2021). Additionally, liquid or fuel
containers based cases are addressed in (Cornillier et al. 2008), (Asawarungsaengkul et al. 2013),
(Lahyani et al. 2015), (Febriandini, Sutopo, et al. 2020), (Wang, Kinable, and Van Woensel 2020),
(Chowmali and Sukto 2021), (Yindong, Liwen, and Jingpeng 2021), (Ramadhani, Masruroh, and
Waluyo 2021) and (Guo et al. 2022). Lastly, studies such as (Henke, Speranza, and Wäscher 2015),
(Gajpal et al. 2017), (Mofid-Nakhaee and Barzinpour 2019) and (Paksaz, Salamian, and Jolai 2020)
focused on garbage collection-related issues.

In conclusion, the MCVRP is a challenging optimization problem that has gained significant attention
in the literature (Ostermeier et al. 2021). Despite the numerous studies and advancements made in
recent years, the MCVRP continues to be an active area of research, with ongoing efforts to develop
efficient and effective solution methods as we analyzed in the prior literature on MCVRP is being
summarized in following table 1.

(Salhi 1987) pioneered MTVRP by conducting multiple trips in the context of vehicle fleet mix.
Within a refinement process, a matching algorithm is proposed to assign routes to vehicles that
are limited to two trips. (Taillard, Laporte, and Gendreau 1996) proposed a three-phase heuristic
algorithm for the MTVRP, which is based on the classical VRP. In the first phase, tabu search is used
to generate a population of routes that satisfy the capacity constraint; in phase two, a set of different
VRP solutions is obtained. In the final one, routes are assigned to vehicles by solving the bin-packing
problem (BPP). Furthermore, in their study, a set of classical MTVRP instances are generated, which
are widely used as benchmarks in the literature. (Brandao and Mercer 1997) investigated a real-world
application of MTVRP with time windows and a heterogeneous fleet and solved the problem using
a tabu search algorithm. The methodology developed in this study was adapted by (Brandão and
Mercer 1998), who solved and compared classical MTVRP instances.

The VRPB deals with the efficient routing of vehicles from a depot to a set of customers, with
the constraint of collecting and delivering goods in the process. The VRPB differs from the classical
Vehicle Routing Problem in that it requires vehicles to return to the depot with goods collected from
certain customers. The numerous studies and the advancements made in recent years, the VRPB
continues to be an active area of research, with ongoing efforts to develop efficient and effective
solution methods. Previous research in the area of VRPB has focused on developing efficient and
effective solution algorithms. For example, in a study (WA et al. 2012), a genetic algorithm was
proposed to solve the VRPB, while (Toth and Vigo 1997) developed a branch-and-bound algorithm
to solve the VRPB to optimality. The proposed model adds to the existing body of research by
considering the use of multi-compartment vehicles in the VRPB, making it a valuable contribution
 5

Table 1. Literature reviewed for MCVRP

Study MPa MC MT MD HVRP Backhauls TW Algorithms

√ √ √ √
This Paper Saving Algorithm
√ √
Wang et al. 2022 MOVND
√ √ √
Chang 2022 ALNS
√ √
Yousra and Ahmed 2022
√ √
Chen, Pan, and Yi 2022 HGA
√ √
Guo et al. 2022 ACO with VND
√ √ √
Frank et al. 2021 ALNS
√ √ √
Ramadhani, Masruroh, and Waluyo 2021
√ √
Yindong, Liwen, and Jingpeng 2021 EDA-LF
√ √
Zou, Pan, and Tasgetiren 2021 IG
√ √
Chowmali and Sukto 2021 FJA-ALNS
√ √ √ √
Paksaz, Salamian, and Jolai 2020
√ √
Wang, Kinable, and Van Woensel 2020 ALNS
√ √
Eshtehadi, Demir, and Huang 2020 EALNS
√ √ √
Febriandini, Sutopo, et al. 2020
√ √ √
Kim and Park 2020
√ √ √
Pasha et al. 2020 EA,VNS,TS,SA
√ √ √
Martins et al. 2019 ALNS
√ √
Chen and Shi 2019 HPSO
√ √
Hübner and Ostermeier 2019 LNS
√ √ √
Chen, Liu, and Langevin 2019 ALNS
√ √
Moőd-Nakhaee and Barzinpour 2019 ALNS & WO
√ √
Mirzaei and Wùhlk 2019 B&P
√ √
Ostermeier et al. 2018 LNS
√ √
Yahyaoui et al. 2020 ALNS, GA
√ √
Alinaghian and Shokouhi 2018 ALNS
√ √
Göçmen and Rızvan 2018
√ √
Ostermeier and Hübner 2018 B& B, LNS
√
Gajpal et al. 2017 ACO
√
Henke, Speranza, and Wäscher 2015 VNS
√ √ √
Lahyani et al. 2015 B&C
√ √
Asawarungsaengkul et al. 2013 Saving Algorithm
√ √ √ √
Cornillier et al. 2008 MPH
√ √
Chajakis and Guignard 2003 LR

a. MP: Multi-Period; MC: Multi-Compartment; MT: Multi-Trip; MD: Multi-Depot; HF: Heterogeneous Fleet; TW:
Time Windows

to the field. For more on VRPB we refer readers to (Koç and Laporte 2018).

3. PROBLEM DESCRIPTION
This paper introduces a Multi-Trip Multi-Compartment Vehicle Routing Problem with Backhauls
(MTMCVRPB) for the delivery of heterogeneous demands together to the linehaul customers and
collection of a variety of goods from the backhaul customers (such as suppliers, returning of expired
products and etc.). Split deliveries are not being considered in this paper. So, each demand for a
particular product should be lower than the corresponding compartment capacity. The orders of
the linehaul customers will be loaded from the depot and after collecting from backhaul customers
unloaded to the depot itself, using same vehicle on a delivery route in a such way that all the linehauls
are satisfied first before visiting any backhaul customer. The objective of this paper is to find the
delivery routes by minimizing two types of costs: fixed cost for minimizing the number of vehicles
and routing cost for minimizing the distance (or time) of traveling. The primary activities of this
problem are to meet the demand of linehaul and backhaul customers by reducing the number of
vehicles, utilising the capacity of each compartment and maximizing the number of trips of a vehicle
in a given time period.

The following figure 1 shows the example of the MTMCVRPB, in which there are 20 linehaul
customers denoted by blue dots, 10 backhaul customers denoted by green triangles and one depot
node denoted by red square in the middle of the figure. Each customer has its demand presented
in table 9 and associated coordinates are also mentioned in this table as well. There are three types
of arcs in the example figure, throughout the blue arcs vehicle if performing delivery process by
visiting linehaul customers, during green arcs vehicle is performing pickups by visiting backhaul
customers and orange arc represents that vehicle is empty at that time.

Figure 1. Example of MTMCVRPB

The following assumptions and constraints govern the characteristics of the customer order and
vehicle routing:
• Every customer should be visited only once in a period.
• Weight or volume of each compartment at vehicle should not exceed its limited capacity.
• One compartment can transport one type of product at a time.
• The loading and unloading time of orders are assumed to be constant and negligible as time
constraints are out of scope of this paper.
• Each compartment should be empty at vehicle before visiting any backhaul customer.
• Other costs, if occur except fixed cost of using vehicles and routing cost of covering distance are
out of scope of this paper too.
• Customer demands may consist of one, or more than one types of items.
• Order of one customer must be delivered by exactly one vehicle as no split delivery is allowed.
The products in a customer’s order must therefore be loaded in the proper compartment of the
same vehicle.
• During a single trip from the depot, each vehicle can deliver to one or more customers.
 7

• It is believed that each customer’s order will be smaller compared to the weight and volume
limitations of each truck. Additionally, it is expected that each item in a customer’s order will be
light in weight and tiny in volume compared to the weight and volume of the compartment on
which it will be loaded.
• The products are designed in such a way that an innumerable number of items can be loaded
into any space of the vehicle, utilizing the full capacity of the space or, at worst, with a negligible
loss in capacity.
• Vehicle can not go directly to backhaul customers but it can return directly from any linehaul if
the demand of all backhaul customers is already fulfilled.
• The quantities to be delivered and picked up are fixed and known in advance.
• Fixed costs will be taken into account for vehicles means does not depend on trips, therefore if a
single vehicle is used for multiple trips, fixed costs will only be taken into account once.

4. MODEL FORMULATION
The MTMCVRPB is defined on an undirected graph G = (V, A) where vertex V represents the set
of n linehaul customers {1, 2, ..., n} and m backhaul customers {n+1, n+2, ..., n+m} and a single
depot {0}. The corresponding set of arcs is denoted by A = {(i, j)| i,j ∈ V: i ̸= j} and a distance dij is
associated with each arc (i, j) ∈ A. All distances are considered asymmetric. Further notations are as
follows:

Index Sets
{0} Depot Node
I, J Set of Customers
K Set of Vehicles
R Set of Trips
L Set of Compartments at Vehicles
Parameters
qlj Demand of product transported through compartment l at customer j
dij Distance from customer i to customer j
Fk Fixed cost of using vehicle k
Ck Travelling cost per km using vehicle k
Qk l Capacity of compartment l at vehicle k
MDT Maximum distance a vehicle travels in a single trip
MDP Maximum distance a vehicle can travel in its all trips
N Number of linehaul customers
M Number of backhaul customers
Decision Variables
Xijkr binary variable having value 1 if vehicle k in its trip r travels from node i to node j
and 0 otherwise
l
Yijkr Flow of delivery products transporting through compartment l of vehicle k in
its trip r at arc i to j
l
Zijkr Flow of pickup products transporting through compartment l of vehicle k in
its trip r at arc i to j
Objective Function

XX
N+M K N+M XX
X N+M K X
R
Min TTC = Fk ∗ X0jk1 + dij ∗ Ck ∗ Xijkr (1)
j=1 k=1 i=0 j=0 k=1 r=1
Basic VRP Constraints

Xijkr = 0 ∀ i = j = 0, 1, 2, ..., N + M; k = 1, 2, ..., K; r = 1, 2, ..., R (2)

X
N
X0jkr ≤ 1 ∀ k = 1, 2, ..., K; r = 1, 2, ..., R; (3)
j=1

X
R
X0jkr ≤ 1 ∀ j = 1, 2, ..., N; k = 1, 2, ..., K; (4)
r=1

X
N+M
Xi0kr ≤ 1 ∀ k = 1, 2, ..., K; r = 1, 2, ..., R (5)
i=1

X
R
Xi0kr ≤ 1 ∀ i = 1, 2, ..., N + M; k = 1, 2, ..., K; (6)
r=1
K X
X R
Xijkr ≤ 1 ∀ i, j = 0, 1, 2, ..., N + M; (7)
k=1 r=1

X
N+M X
N+M
Xijkr – Xjikr = 0 ∀ j = 1, 2, ..., N + M; k = 1, 2, ..., K; r = 1, 2, ..., R; (8)
i=0 i=0

XX
N+M K X
R
Xijkr = 1 ∀ i = 1, 2, ..., N + M; (9)
j=1 k=1 r=1

N+M X
X N+M X N+M
N+M X
Xijkr – Xijk(r–1) ≤ 0 ∀ k = 1, 2, ..., K; r = 2, 3, ..., R (10)
i=0 j=0 i=0 j=0

N X
X R X
R
Xijkr ≥ X1jkr ∀ j = 0, 1, 2, ..., N + M; k = 1, 2, ..., K (11)
i=0 r=1 r=1

Balancing Flow Constraints

A. Flow control of linehaul customers

K X
N X
X R X
L X K X
N X R X
L
l l
Yijkr – Yjikr = qlj ∀ j = 1, 2, ..., N; (12)
i=0 k=1 r=1 l=1 i=0 k=1 r=1 l=1

qlj ∗ Xijkr ≤ Yijkr

l
∀ i, j = 0, 1, 2, ..., N; k = 1, 2, ..., K; r = 1, 2, ..., R (13)
l
Yijkr ≤ (Qkl – qli ) ∗ Xijkr ∀ i, j = 0, 1, 2, ..., N; k = 1, 2, ..., K; r = 1, 2, ..., R (14)

B. Flow control of backhaul customers

X X
N+M K X
R XX
N+M K X
R X X
N+M K X
R
l l
Zjikr – Zjikr = qlj ∀ j = N + 1, N + 2, ..., N + M; l = 1, 2, ..., L
i=0,N k=1 r=1 i=0 k=1 r=1 i=0,N k=1 r=1
(15)
 9

qlj ∗ Xijkr ≤ Zijkr

l
∀ i = 0, 1, 2, ..., N +M; j = 0, N +1, N +2, ..., N +M; k = 1, 2, ..., K; r = 1, 2, ..., R (16)
l
Zijkr ≤ (Qkl –qli ) ∗ Xijkr ∀ i = 0, 1, 2, ..., N +M; j = 0, N +1, N +2, ..., N +M; k = 1, 2, ..., K; r = 1, 2, ..., R
(17)
Maximum distance allowed for per trip

X N+M
N+M X
dij ∗ Xijkr ≤ MDT ∈ {0, 1} ∀ k = 1, 2, ..., K; r = 1, 2, ..., R (18)
i=0 j=0

Maximum distance allowed in all trips

N+M XX
X N+M R
dij ∗ Xijkr ≤ MDP ∈ {0, 1} ∀ k = 1, 2, ..., K; r = 1, 2, ..., R (19)
i=0 j=0 r=1

Restrictions on Decision Variables

Xijkr ∈ {0, 1} ∀ i, j = 0, 1, 2, ..., N + M; k = 1, 2, ..., K; r = 1, 2, ..., R (20)

l l
Yijkr , Zijkr ≥ 0 ∀ i, j = 0, 1, 2, ..., N + M; k = 1, 2, ..., K; r = 1, 2, ..., R; l = 1, 2, ...L (21)

The objective function of the formulation, denoted as (1), balances two types of costs: the fixed
cost of using vehicles and the traveling cost of routing them. The constraints, (2) to (21), serve to
govern the various aspects of the vehicle routing process. Constraint (2) ensures that the vehicles
do not follow a looping route except depot node, while (3) and (4) restrict that each vehicle make
at most one departure from the depot to any linehaul customer in one trip. Similarly, (5) and (6)
dictate that each vehicle can only enter the depot once after visiting all customers on its route. The
constraint (7) ensures that a single vehicle during its any trip can visit at most one arc. (8) requires
that if a vehicle enters a node (excluding the depot) during its r th trip, it must also depart from that
node. The constraint (9) ensures that every node is visited exactly once by a single vehicle during
its entire trip. The constraints (10) and (11) dictate that another trip by any vehicle can only start
after the completion of the previous trip. Meanwhile, (12), (13) and (14) balance the flow of linehaul
customers, while (15), (16) and (17) regulate the flow of backhaul customers. The total distance
covered by vehicles during their trips is controlled by constraints (18) and (19). Finally, constraint
(20) restricts the decision variables to binary values and (21) requires that the vehicles must carry a
positive flow.

5. SOLUTION APPROACH
Heuristics are problem-solving techniques that use practical, rule-of-thumb strategies to find solu-
tions that are not guaranteed to be optimal but are usually good enough for practical purposes. A
saving-based heuristic is a type of heuristic that is commonly used in vehicle routing problems.

In a saving-based heuristic, the algorithm seeks to minimize the total distance or cost of a set
of routes that visit a given set of locations. The heuristic works by first computing the distance
between all pairs of locations and then ordering the pairs in decreasing order of the savings achieved
by combining the two locations into a single route. The algorithm then iteratively merges the pairs
of locations with the highest savings until all locations are visited.

By using a saving-based heuristic, you can quickly find near-optimal solutions to vehicle rout-
ing problems, without having to search through all possible routes. However, it’s important to note
that heuristics like this are not guaranteed to find the optimal solution and there may be trade-offs
between solution quality and computational efficiency.

The central idea of this algorithm is to optimize the savings achieved by consolidating two routes
and having them serviced by a single vehicle, instead of using two different vehicles. To achieve
this, the algorithm considers two distinct routes, represented by first route and second route and
identifies the last customer on first route and the first customer on second route, denoted as i and
j, respectively. By combining these routes and having them serviced by a single vehicle, the total
distance that the vehicle needs to travel decreases by a value equal to Sij , which can be computed
using the following function:
Sij = di0 + d0j – dij
(Toth and Vigo 2002) provides a detailed description of the algorithm as well as an analysis of its
performance on a variety of test problems. The authors also discuss several extensions and variations
of the algorithm, including a modified version that incorporates a time window constraint.

Here is a step-wise description of a saving-based heuristic for the proposed MTMCVRPB:

Step 1: Compute the savings for each pair of locations in the problem.
Step 2: Sort the savings in descending order.
Step 3: Initialize a set of routes, where each route starts and ends at the depot.
Step 4: Take the pair of locations with the highest savings and attempt to merge them into a single
route. To do this, identify the routes that include the two locations and determine the best way to
combine them. There are several possible ways to merge the routes, such as inserting the second
route into the first route, reversing one of the routes and concatenating them, or creating a new
route that connects the two locations. Also take care of the basic assumptions of VRPBs. Choose the
option that minimizes the total distance traveled.
Step 5: Update the set of routes with the merged route and remove the original routes that were
combined.
Step 6: Repeat steps 4 and 5 until all locations are visited.
If necessary, repeat steps 1 to 6 with a different set of initial routes or parameter settings.
This saving-based heuristic can provide near-optimal solutions for large-scale problems in a relatively
short amount of time.

6. COMPUTATIONAL EXPERIMENTATION
To evaluate the effectiveness of the proposed mixed integer linear programming (MILP) model for
the Multi-Trip Multi-Compartment Vehicle Routing Problem with Backhauls (MTMCVRPB), we
conducted numerical experiments using randomly generated data sets.

Here, we present the results of the numerical experiments conducted on a simple example us-
ing LINGO 19. In this example we assumed that there is a single depot and 10 customers with
two types of products and three vehicles (heavy, medium and light) each having two equal sized
compartments with different capacities. The vehicles have fixed costs and per km costs associated
with them. All the attributes of vehicles in table 10 are presumed for the sake of experimentational
purposes. The demand for both items from each customer is given in table 8 and a symmetric distance
matrix that was produced randomly is included in table 7. Each vehicle may make a maximum of
two trips, each lasting 250 kilometers.

The experiment utilized different vehicles, the assumed details of them are presented in table 10.
The vehicles are categorized as heavy, medium, and light, with fixed costs of Rs. 1000, Rs. 800, and
Rs. 600, respectively. In terms of per-kilometer costs, the heavy vehicle cost Rs. 35 per km, the
 11

medium vehicle cost Rs. 30 per km, and the light vehicle cost Rs. 25 per km. Each vehicle have two
compartments, with a capacity of 300 units for both compartments in the case of heavy vehicles, 200
units for both compartments in the case of medium vehicles, and 100 units for both compartments in
the case of light vehicles.

We solve the problem using the proposed MILP model and a saving-based heuristic. We eval-
uate the built-in model against a simple example using the precise branch and bound approach
via LINGO and model validation is carried out to ensure that the produced model is working in
accordance with the anticipated logic or by replicating the behavior of the system.

By varying the linehaul customers (LC) and backhaul customers (BC), we find the following table 2
of solutions in which total cost (TC) is the summation of fixed cost (FC) and routing cost (RC). Note
that after any route, values in square brackets represents the vehicle name and then its trip number
e.g. optimal route for instance 1 in table 1 is Depot-C8-C10-C3-C1-C4-C2-Depot (Medium,1);
Depot-C7-C9-C5-C6-Depot (Medium,2), here (Medium,1) represents that vehicle 2 travels on route
Depot-C8-C10-C3-C1-C4-C2-Depot during its 1st trip and (Medium,2) represents that vehicle
2 travel on route Depot-C7-C9-C5-C6-Depot in its 2nd trip. For the instances in which LC = 2
and LC = 1, the solution for them is infeasible as no solution is being found by the solver. So, these
instances are not included in table 2.

Sensitivity analysis is an important aspect of optimization problems to examine the behavior of

the proposed model under different scenarios. In this regard, we conducted sensitivity analysis on the
maximum distance per trip (MDT), maximum distance per period for vehicle (MDP), fixed costs of
each vehicle (Fk ), cost per km for each vehicle (Ck ), maximum number of trips (R) that each vehicle
can perform and variation in the number of products or number of compartments (L).

For the instance of 6 linehauls and 4 backhauls, if we changes distance parameter MDT from
350 to ∞ then the solution will remain unchanged and if we make changes in MDT between 0 and
350 then the total cost changes as shown in figure 2.

The following table 3 shows the impact of MDT on number of trips of vehicles as you can see
here MDT has reverse impact on total cost means if MDT decreases total cost will be increased. We
also shown through table 5 that if we decrease MDP, FC may decrease but RC increases as MDP
controls the number of trips a vehicle can perform during whole period.

The impact of changing the travelling cost per km parameter, Ck , in the objective function ?? on
the outcome of instance 5 of table 2 is shown in figure 5, which demonstrates that changes in the
cost per km of the medium vehicle have the greatest influence on the solution when compared to
the heavy vehicle and light vehicle. We also see that if we changes Ck greater than 30 for all three
vehicles, the total cost is approximately equal and unchanged.

In order to examine the effect of changes in the fixed cost parameter Fk on the solution of in-
stance 5 presented in table 2, a sensitivity analysis was performed by varying the values of Fk in
the objective function 1. The results of this analysis are displayed in figure 6. It was observed that
modifications made to Fk of heavy vehicles had the least impact on the solution as compared to
changes made to the Fk of medium and light vehicles. The model in this example exhibits high
sensitivity to medium vehicle, as the total cost consistently increases with the associated fixed cost.

Varying the number of trips for the instances in table 1, we see the variation in total cost as shown
in figure 3. If number of trips increases, total cost either will remain same for some instances or it
decreases for other instances.

We also tested the sensitivity of the proposed mathematical model against the number of prod-
ucts or number of compartments involved in the experiments. In this experiment also, we used
instance in which LC = 6 and BC = 4 i.e. instance 5 in table 2. As you can see in the table 4 we varied
the number of products from 1 to 5 and clearly can be seen that there is very less impact on fixed cost
and number of routes but routing cost gradually increases as number of products increase and this
impact graphically can be understood by the figure 4 in this figure green portion represents fixed
cost and blue portion of the column represents routing cost and whole column represents the total cost.

As seen from the table 6, it is evident that there is significant deviation in the optimal solution
and solution provided by the proposed algorithm i.e. saving based heuristic but the algorithm finds
the near optimal solution very quickly.

Table 2. Different instances and solution routes

Instance LC BC FC RC TC Optimal Routes

Depot-C8-C10-C3-C1-C4-C2-Depot (Medium,1);
1 10 0 800 12435 13235
Depot-C7-C9-C5-C6-Depot (Medium,2)
Depot-C2-C4-C1-C3-C6-C10-Depot (Medium,1);
2 9 1 800 12093.6 12893.6
Depot-C8-C5-C9-C7-Depot (Medium,2)
Depot-C2-C4-C1-C3-C6-C10-Depot (Medium,1);
3 8 2 800 12569.4 13369.4
Depot-C8-C5-C7-C9-Depot (Medium,2)
Depot-C4-C1-C3-C6-C8-C10-Depot (Medium,1);
4 7 3 800 12167.4 12967.4
Depot-C2-C7-C5-C9-Depot (Medium,2)
Depot-C3-C6-C8-C10-Depot (Medium,1);
5 6 4 1400 11233.5 12633.5 Depot-C2-C4-C1-Depot (Light,1);
Depot-C5-C9-C7-Depot (Light,2)
Depot-C2-C4-C1-C3-C10-C6-C8-Depot (Medium,1);
6 5 5 1400 10781.75 12181.75
Depot-C5-C9-C7-Depot (Light,1)
Depot-C4-C1-C3-C10-C6-C8-Depot (Medium,1);
7 4 6 800 11343.9 12143.9
Depot-C2-C7-C9-C5-Depot (Medium,2)
Depot-C2-C7-C9-C5-C8-C6-C10-Depot (Heavy,1);
8 3 7 1600 11072.2 12672.2
Depot-C3-C1-C4-Depot (Light,1)
Depot-C2-C7-C9-C5-C8-Depot (Medium,1);
9 2 8 800 12543.6 13343.6
Depot-C1-C4-C3-C10-C6-Depot (Medium,2)
 13

Figure 2. Impact of maximum route length on total cost

Figure 3. Variations in total cost based on variations in the maximum number of trips (R)

Figure 4. Impact on Total Cost of Variations in Number of Products

Figure 5. Impact on Total Cost of Variations in Cost per KM

 Figure 6. Impact on Total Cost of Variations in Fixed Cost per Vehicle

Table 3. Impact of MDT on trips of vehicles

MDT FC RC TC Optimal Routes

Depot-C6-C8-C10-Depot (Medium,1);
Depot-C3-C1-Depot (Medium,2);
150 1400 12623.1 14023.1
Depot-C5-C9-C7-Depot (Light,1);
Depot-C2-C4-Depot (Light,2)
Depot-C2-C3-C6-C8-C10-Depot (Medium,1);
175 1400 12179.9 13579.9 Depot-C5-C9-C7-Depot (Light,1);
Depot-C1-C4-Depot (Light,2)
Depot-C3-C6-C8-C10-Depot (Medium,1);
200 1400 11233.5 12633.5 Depot-C2-C4-C1-Depot (Light,1);
Depot-C5-C9-C7-Depot (Light,2)
Depot-C3-C6-C8-C10-Depot (Medium,1);
250 1400 11233.5 12633.5 Depot-C2-C4-C1-Depot (Light,1);
Depot-C5-C9-C7-Depot (Light,2)
Depot-C2-C4-C1-C3-C6-C8-C10-Depot (Medium,1);
350 1400 11150.8 12550.8
Depot-C5-C9-C7-Depot (Light,1)
Depot-C2-C4-C1-C3-C6-C8-C10-Depot (Medium,1);
∞ 1400 11150.8 12550.8
Depot-C5-C9-C7-Depot (Light,1)

Table 4. Impact on Total Cost making variations in number of products

Products Fixed Cost Routing Cost Total Cost Number of Routes Routes
Depot-C2-C4-C1-C3-C10-C8-Depot (Light,1);
1 600 10362.5 10962.5 2
Depot-C6-C5-C9-C7-Depot (Light,2)
Depot-C3-C6-C8-C10-Depot (Medium, 1);
2 1400 11233.55 12633.55 3 Depot-C2-C4-C1-Depot (Light,1);
Depot-C5-C9-C7-Depot (Light,2)
Depot-C2-C4-C1-C3-Depot (Medium,1);
3 1400 11482.8 12882.8 3 Depot-C6-C8-C10-Depot (Light,1);
Depot-C5-C9-C7-Depot (Light,2)
Depot-C3-C6-C8-C10-Depot (Medium,1);
4 1400 12150.15 13550.15 3 Depot-C1-C4-C2-Depot (Medium,2);
Depot-C5-C9-C7-Depot (Light,1)
Depot-C4-C2-C5-C9-C7-Depot (Medium,1);
5 800 12558.3 13358.3 2
Depot-C1-C3-C6-C8-C10-Depot (Medium,2)
 15

Table 5. Impact of MDP on Fixed Cost and Routing Cost

MDP MDT FC RC TC Optimal Routes

Depot-C6-C8-C10-Depot (Medium,1);
Depot-C3-C1-Depot (Medium,2);
>=300 150 1400 12623.1 14023.1
Depot-C5-C9-C7-Depot (Light,1);
Depot-C2-C4-Depot (Light,2)
Depot-C4-C2-Depot (Medium,1);
Depot-C3-C1-Depot (Medium,2);
250 150 1400 12739.5 14139.5
Depot-C6-C8-C10-Depot (Light,1);
Depot-C5-C9-C7-Depot (Light,2)
Depot-C1-C3-Depot (Heavy,1);
Depot-C5-C9-C7-Depot (Medium,1);
225 150 2400 13403.8 15803.8
Depot-C6-C8-C10-Depot (Light,1);
Depot-C4-C2-Depot (Light,2)
Depot-C4-C2-Depot (Heavy,1);
Depot-C5-C9-C7-Depot (Medium,1);
200 150 2400 14507.1 16907.1 Depot-C3-Depot (Medium,2);
Depot-C6-C8-C10-Depot (Light,1);
Depot-C1-Depot (Light,2)
Depot-C5-C9-C7-Depot (Heavy,1);
Depot-C3-C10-Depot (Medium,1);
190 150 2400 14831.4 17231.4 Depot-C2-C5-Depot (Medium,2);
Depot-C6-C8-Depot (Light,1);
Depot-C1-Depot (Light,2)
Depot-C3-C6-C8-C10-Depot (Heavy,1);
175 175 2400 12888.1 15288.1 Depot-C2-C5-C9-C7-Depot (Medium,1);
Depot-C1-C4-Depot (Light,1)
Depot-C2-C4-Depot (Medium,1);
Depot-C1-C3-Depot (Medium,2);
250 175 1400 12739.5 14139.5
Depot-C6-C8-C10-Depot (Light,1);
Depot-C5-C9-C7-Depot (Light,2)
Depot-C2-C5-C9-C7-Depot (Medium,1);
300 175 1400 12339.5 13739.5 Depot-C3-C6-C8-C10-Depot (Medium,2);
Depot-C4-C1-Depot (Light,1)
Depot-C2-C3-C6-C8-C10-Depot (Medium,1);
>=350 175 1400 12179.9 13579.9 Depot-C5-C9-C7-Depot (Light,1);
Depot-C1-C4-Depot (Light,2)
Depot-C3-C6-C8-C10-Depot (Medium,1);
>=350 200 1400 11233.5 12633.5 Depot-C2-C4-C1-Depot (Light,1);
Depot-C5-C9-C7-Depot (Light,2)
Depot-C3-C1-C4-C2-Depot (Medium,1);
300 200 1400 11482.8 12882.8 Depot-C5-C9-C7-Depot (Light,1);
Depot-C6-C8-C10-Depot (Light,2)
Depot-C6-C8-C10-Depot (Heavy,1);
200 200 2400 12432.1 14832.1 Depot-C2-C4-C1-C3-Depot (Medium,1);
Depot-C5-C9-C7-Depot (Light,1)
Depot-C3-C1-C4-C2-Depot (Medium,1);
250 250 1400 11482.8 12882.8 Depot-C5-C9-C7-Depot (Light,1);
Depot-C6-C8-C10-Depot (Light,2)
Depot-C3-C6-C8-C10-Depot (Medium,1);
>=300 250 1400 11233.5 12633.5 Depot-C2-C4-C1-Depot (Light,1);
Depot-C5-C9-C7-Depot (Light,2)
Depot-C2-C4-C1-C3-C6-C8-C10-Depot (Medium,1);
>=350 >=350 1400 11150.8 12550.8
Depot-C5-C9-C7-Depot (Light,1)
 Table 6. Computational time and total cost using LINGO 19 vs saving algorithm

Optimal Cost Cost Using Time Time Using

LC BC Using LINGO Saving Algorithm Using LINGO Saving Algorithm
(in Rs.) (in Rs.) (in seconds) (in seconds)
10 0 13235 17550.25 41.26 0.003995
9 1 12893.6 17621.15 4.83 0.003001
8 2 13369.4 18564.30 2.17 0.003994
7 3 12967.4 16646.75 1.65 0.003998
6 4 12633.5 16947.85 1.69 0.002998
5 5 12181.75 16129.2 2.37 0.002998

7. CONCLUSION
This research paper presents a novel approach to Multi-Trip MCVRP with backhauls, aiming to
increase the utilization of multi-compartment vehicles and minimizing the total cost of transportation
including fixed cost and routing cost. To achieve this objective, we incorporated a maximum distance
constraint for each vehicle’s route as vehicle may have limited fuel tank capacity and other similar
restrictions, also we included fixed cost in the objective function to reduce the number of vehicles
required.

The proposed model was solved using LINGO 19.0 software and a saving-based heuristic algorithm
was proposed for a quick solution. To evaluate the benefits of vehicle selection, we conducted a series
of numerical experiments on randomly generated data, which demonstrated the significance of our
approach in various scenarios.

The results show that the proposed MILP model is effective in solving the MTMCVRPB and
can be used to optimize the routes for a heterogeneous fleet of multi-compartment vehicles to
transport heterogeneous commodities simultaneously. The saving-based heuristic can be used to
find quick solutions to the problem. The proposed model can be beneficial for industries such as
grocery and food delivery, fuel distribution and garbage collection.

The practical implications of our study are substantial, as the proposed model can be utilized by
logistics companies to optimize delivery routes, reduce transportation costs and enhance delivery
efficiency. For researchers, this study introduces a new kind of VRP model with distance constraints,
providing new avenues for further research in the field.

However, some limitations of the study include the assumption of fixed and known customer demand
and the absence of time constraints. Future research can address these limitations by exploring
methods for handling dynamic demand and incorporating time-based constraints.

In conclusion, this article makes a significant addition to the field of VRP by introducing a novel
approach for the Multi-Trip MCVRP with backhauls, providing useful implications for logistics
firms, and laying the foundation for further study in the area.
 17

8. COMPLIANCE WITH ETHICAL STANDARDS

Disclosure of potential conflicts of interest: The authors declare that there is no potential conflict
of interest.

Research involving Human Participants and/or Animals: Not Applicable.

Informed consent: Not Applicable.

References
Alinaghian, Mahdi, and Nadia Shokouhi. 2018. Multi-depot multi-compartment vehicle routing problem, solved by a hybrid
adaptive large neighborhood search. Omega 76:85–99.
Asawarungsaengkul, K, Tarit Rattanamanee, and Teeradej Wuttipornpun. 2013. A multi-size compartment vehicle routing
problem for multi-product distribution: models and solution procedures. International Journal of Artificial Intelligence 11
(A13): 237–256.
Brandao, José, and Alan Mercer. 1997. A tabu search algorithm for the multi-trip vehicle routing and scheduling problem.
European journal of operational research 100 (1): 180–191.
Brandão, José Carlos Soares, and Alan Mercer. 1998. The multi-trip vehicle routing problem. Journal of the Operational research
society 49:799–805.
Cattaruzza, Diego, Nabil Absi, and Dominique Feillet. 2016. The multi-trip vehicle routing problem with time windows and
release dates. Transportation Science 50 (2): 676–693.
Chajakis, Emmanual D, and Monique Guignard. 2003. Scheduling deliveries in vehicles with multiple compartments. Journal
of Global Optimization 26 (1): 43–78.
Chang, Peihuan. 2022. The vehicle routing optimization problem of refrigerated vehicle with temperature compartments
under consistent delivery model. Academic Journal of Science and Technology 1 (2): 1–9.
Chen, Jiumei, and Jing Shi. 2019. A multi-compartment vehicle routing problem with time windows for urban distribution–a
comparison study on particle swarm optimization algorithms. Computers & Industrial Engineering 133:95–106.
Chen, Liyin, Fubin Pan, and Junmin Yi. 2022. Multi-compartment vehicle routing optimization based on a hybrid genetic
algorithm. 2303 (1): 012042.
Chen, Lu, Yang Liu, and André Langevin. 2019. A multi-compartment vehicle routing problem in cold-chain distribution.
Computers & Operations Research 111:58–66.
Chowmali, W, and S Sukto. 2021. A hybrid fja-alns algorithm for solving the multi-compartment vehicle routing problem
with a heterogeneous fleet of vehicles for the fuel delivery problem. Decision Science Letters 10 (4): 497–510.
Clarke, Geoff, and John W Wright. 1964. Scheduling of vehicles from a central depot to a number of delivery points. Operations
research 12 (4): 568–581.
Coelho, Leandro C, and Gilbert Laporte. 2015. Classification, models and exact algorithms for multi-compartment delivery
problems. European Journal of Operational Research 242 (3): 854–864.
Cornillier, Fabien, Fayez F Boctor, Gilbert Laporte, and Jacques Renaud. 2008. A heuristic for the multi-period petrol station
replenishment problem. European Journal of Operational Research 191 (2): 295–305.
Dantzig, George B, and John H Ramser. 1959. The truck dispatching problem. Management science 6 (1): 80–91.
Derigs, Ulrich, Jens Gottlieb, Jochen Kalkoff, Michael Piesche, Franz Rothlauf, and Ulrich Vogel. 2011. Vehicle routing with
compartments: applications, modelling and heuristics. OR spectrum 33:885–914.
Eshtehadi, Reza, Emrah Demir, and Yuan Huang. 2020. Solving the vehicle routing problem with multi-compartment
vehicles for city logistics. Computers & Operations Research 115:104859.
Febriandini, IF, W Sutopo, et al. 2020. Multi-compartment vehicle routing problem to find the alternative distribution route
of petroleum product delivery. 943 (1): 012039.
Fleischmann, Bernhard. 1990. The vehicle routing problem with multiple use of vehicles. Fachbereich Wirtschaftswissenschaften,
Universität Hamburg.
Frank, Markus, Manuel Ostermeier, Andreas Holzapfel, Alexander Hübner, and Heinrich Kuhn. 2021. Optimizing routing
and delivery patterns with multi-compartment vehicles. European Journal of Operational Research 293 (2): 495–510.
Gajpal, Yuvraj, MMS Abdulkader, Shuai Zhang, and SS Appadoo. 2017. Optimizing garbage collection vehicle routing
problem with alternative fuel-powered vehicles. Optimization 66 (11): 1851–1862.
Göçmen, Elifcan, and EROL Rızvan. 2018. Location and multi-compartment capacitated vehicle routing problem for blood
banking system. International Journal of Engineering Technologies IJET 4 (1): 1–12.
Goetschalckx, Marc, and Charlotte Jacobs-Blecha. 1989. The vehicle routing problem with backhauls. European Journal of
Operational Research 42 (1): 39–51.
Golden, BL, E Baker, J Alfaro, and J Schaffer. 1985. The vehicle routing problem with backhauling: two approaches. In
Proceedings of the twenty-first annual meeting of the se tims, myrtle beach, sc, usa.
Guo, Ning, Bin Qian, Jing Na, Rong Hu, and Jian-Lin Mao. 2022. An enhanced ant colony algorithm with variable
neighborhood descent for multi-compartment vehicle routing problem with time limits, 1961–1966.
Henke, Tino, M Grazia Speranza, and Gerhard Wäscher. 2015. The multi-compartment vehicle routing problem with flexible
compartment sizes. European Journal of Operational Research 246 (3): 730–743.
Hübner, Alexander, and Manuel Ostermeier. 2019. A multi-compartment vehicle routing problem with loading and unloading
costs. Transportation Science 53 (1): 282–300.
Kim, Heejun, and Taehyung Park. 2020. Distribution scheduling of multi-temperature compartments delivery vehicles. ICIC
express letters. Part B, Applications: an international journal of research and surveys 11 (6): 559–565.
Koç, Çağrı, and Gilbert Laporte. 2018. Vehicle routing with backhauls: review and research perspectives. Computers &
Operations Research 91:79–91.
Koch, Henriette, Tino Henke, and Gerhard Wäscher. 2016. A genetic algorithm for the multi-compartment vehicle routing
problem with flexible compartment sizes. Working Paper Series.
Kurnia, Hari, Elyza Gustri Wahyuni, Elang Cergas Pembrani, Syifa Tri Gardini, and Silfa Kurnia Aditya. 2018. Vehicle
routing problem using genetic algorithm with multi compartment on vegetable distribution. In Iop conference series:
materials science and engineering, 325:012012. 1. IOP Publishing.
Lahyani, Rahma, Leandro C Coelho, Mahdi Khemakhem, Gilbert Laporte, and Frédéric Semet. 2015. A multi-compartment
vehicle routing problem arising in the collection of olive oil in tunisia. Omega 51:1–10.
Martins, Sara, Manuel Ostermeier, Pedro Amorim, Alexander Hübner, and Bernardo Almada-Lobo. 2019. Product-oriented
time window assignment for a multi-compartment vehicle routing problem. European Journal of Operational Research
276 (3): 893–909.
Mingozzi, Aristide, Roberto Roberti, and Paolo Toth. 2013. An exact algorithm for the multitrip vehicle routing problem.
INFORMS Journal on Computing 25 (2): 193–207.
Mirzaei, Samira, and Sanne Wøhlk. 2019. A branch-and-price algorithm for two multi-compartment vehicle routing problems.
EURO Journal on Transportation and Logistics 8 (1): 1–33.
Mofid-Nakhaee, Elham, and Farnaz Barzinpour. 2019. A multi-compartment capacitated arc routing problem with interme-
diate facilities for solid waste collection using hybrid adaptive large neighborhood search and whale algorithm. Waste
Management & Research 37 (1): 38–47.
Mor, Andrea, and Maria Grazia Speranza. 2022. Vehicle routing problems over time: a survey. Annals of Operations Research
314 (1): 255–275.
Ostermeier, Manuel, Tino Henke, Alexander Hübner, and Gerhard Wäscher. 2021. Multi-compartment vehicle routing
problems: state-of-the-art, modeling framework and future directions. European Journal of Operational Research 292 (3):
799–817.
Ostermeier, Manuel, and Alexander Hübner. 2018. Vehicle selection for a multi-compartment vehicle routing problem.
European Journal of Operational Research 269 (2): 682–694.
Ostermeier, Manuel, Sara Martins, Pedro Amorim, and Alexander Hübner. 2018. Loading constraints for a multi-compartment
vehicle routing problem. Or Spectrum 40 (4): 997–1027.
Paksaz, Amir Mohammad, Farima Salamian, and Fariborz Jolai. 2020. Waste collection problem with multi-compartment
vehicles and fuzzy demands.
 19

Pasha, Junayed, Maxim A Dulebenets, Masoud Kavoosi, Olumide F Abioye, Hui Wang, and Weihong Guo. 2020. An
optimization model and solution algorithms for the vehicle routing problem with a “factory-in-a-box”. Ieee Access
8:134743–134763.
Ramadhani, Dinda Safitri, Nur Aini Masruroh, and Joko Waluyo. 2021. Model of vehicle routing problem with split delivery,
multi trips, multi products and compartments for determining fuel distribution routes. ASEAN Journal of Systems
Engineering 5 (2): 51–55.
Reil, Sebastian, Andreas Bortfeldt, and Lars Mönch. 2018. Heuristics for vehicle routing problems with backhauls, time
windows, and 3d loading constraints. European Journal of Operational Research 266 (3): 877–894.
Salhi, S. 1987. The integration of routing into the location-allocation and vehicle composition problems. PhD diss., University
of Lancaster.
Santos, Maria João, Pedro Amorim, Alexandra Marques, Ana Carvalho, and Ana Póvoa. 2020. The vehicle routing problem
with backhauls towards a sustainability perspective: a review. Top 28 (2): 358–401.
Şen, Ahmet, and Kerem Bülbül. 2008. A survey on multi trip vehicle routing problem.
Taillard, Eric D, Gilbert Laporte, and Michel Gendreau. 1996. Vehicle routeing with multiple use of vehicles. Journal of the
Operational research society 47 (8): 1065–1070.
Toth, Paolo, and Daniele Vigo. 1997. An exact algorithm for the vehicle routing problem with backhauls. Transportation
science 31 (4): 372–385.
. 2002. The vehicle routing problem. SIAM.
. 2014. Vehicle routing: problems, methods, and applications. SIAM.
WA, W Nurfahizul Ifwah, M Shaiful, MZ Shamsunarnie, ZM Zainuddin, and M Fuad. 2012. Genetic algorithm for vehicle
routing problem with backhauls. Journal of Science and Technology 4 (1).
Wang, L, Joris Kinable, and Tom Van Woensel. 2020. The fuel replenishment problem: a split-delivery multi-compartment
vehicle routing problem with multiple trips. Computers & Operations Research 118:104904.
Wang, Xuping, Wenping Fan, Hongxin Zhan, and Zilai Sun. 2022. Variable neighborhood descent for multi-compartment
and multi-objective vehicle routing problem in refined product distribution. Asia-Pacific Journal of Operational Research
39 (02): 2150019.
Yahyaoui, Hiba, Islem Kaabachi, Saoussen Krichen, and Abdulkader Dekdouk. 2020. Two metaheuristic approaches for
solving the multi-compartment vehicle routing problem. Operational Research 20 (4): 2085–2108.
Yindong, SHEN, PENG Liwen, and LI Jingpeng. 2021. An improved estimation of distribution algorithm for multi-
compartment electric vehicle routing problem. Journal of Systems Engineering and Electronics 32 (2): 365–379.
Yousra, Bouleft, and Elhilali Alaoui Ahmed. 2022. Multi-compartment vehicle routing problem for waste collection in smart
cities, 1–7.
Zou, Wen-Qiang, Quan-Ke Pan, and M Fatih Tasgetiren. 2021. An effective iterated greedy algorithm for solving a
multi-compartment agv scheduling problem in a matrix manufacturing workshop. Applied Soft Computing 99:106945.
 TABLES
Table 7. Distance between Customers

Distance Depot C1 C2 C3 C4 C5 C6 C7 C8 C9 C10

Depot 0 49.5 19.31 20.81 55.04 34.23 37.7 60.96 32.65 41.11 24
C1 49.5 0 61.33 38.33 62.65 83.05 68.94 105.99 72.28 90.55 51.87
C2 19.31 61.33 0 39.82 43.86 34.83 53.46 44.78 45.22 37.16 42.58
C3 20.81 38.33 39.82 0 67.9 47.8 31.62 81.32 33.96 56.65 13.89
C4 55.04 62.65 43.86 67.9 0 78.52 92.7 69.72 86.95 79.4 77.49
C5 34.23 83.05 34.83 47.8 78.52 0 36.89 45.61 24.7 9.9 39.45
C6 37.7 68.94 53.46 31.62 92.7 36.89 0 81.54 12.21 46.62 17.8
C7 60.96 105.99 44.78 81.32 69.72 45.61 81.54 0 69.64 37.34 78.82
C8 32.65 72.28 45.22 33.96 86.95 24.7 12.21 69.64 0 34.41 21.02
C9 41.11 90.55 37.16 56.65 79.4 9.9 46.62 37.34 34.41 0 49.09
C10 24 51.87 42.58 13.89 77.49 39.45 17.8 78.82 21.02 49.09 0

Table 8. Demand of Linehaul and Backhaul Customers

Demand Depot C1 C2 C3 C4 C5 C6 C7 C8 C9 C10

Product 1 0 9 26 95 56 77 20 82 25 13 29
Product 2 0 17 32 0 2 46 84 30 59 24 30

Table 9. Coordinates and demand of customers associated with the example given in őgure 1

Customers Depot C1 C2 C3 C4 C5 C6 C7 C8 C9 C10 C11 C12 C13 C14 C15 C16 C17 C18 C19 C20 C21 C22 C23 C24 C25 C26 C27 C28 C29 C30
Coordinates x 50 27 75 51 25 0 90 26 40 65 31 13 59 76 39 99 71 8 69 61 24 98 95 15 81 37 9 8 9 33 57
y 50 100 2 97 28 51 27 60 24 93 4 93 100 66 92 48 15 91 28 27 77 60 35 5 2 46 20 5 41 37 79
Demand Product 1 0 29 11 50 26 40 13 16 21 38 4 23 0 16 40 18 16 24 0 37 34 35 43 23 39 32 7 22 6 29 6
Product 2 0 25 27 36 25 46 30 28 42 36 29 50 43 39 30 35 43 20 12 19 48 19 16 23 0 34 3 28 19 3 42

Table 10. Vehicles used in the experiments

Vehicle Heavy Medium Light

Capacity Compartment 1 300 200 100
Compartment 2 300 200 100
Fixed Cost 1000 800 600
Cost per km 35 30 25